Use a set of (continuous) columns to choose a good set of rows

make_metric_df(
  df,
  mets_df = tibble::tibble(metric = "av_clust_size", high_good = TRUE, clust_sum = TRUE,
    level = "clustering"),
  context = c("method", "groups"),
  mets_col = "summary_mets",
  summarise_method = median,
  scale = FALSE,
  top_thresh = 0.25,
  best_thresh = 5,
  level = c("across", "within")
)

Arguments

df: Dataframe with columns over which to find good rows
mets_df: Dataframe mapping the name of possible metrics to cases (columns) in which to use that metric.
context: Character. Name of columns in df that define context.
mets_col: Character. Name of mets_df column to use in this instance.
summarise_method: Character. Name of method to use in summarising if there is more than one row per context.
scale: Logical. If true, all metrics will be rescale 0 ('worst') to 1 ('best').
top_thresh: Numeric specifying the proportion of rows considered 'top'.
best_thresh: Numeric specifying the absolute number of rows considered 'best'.
level: Either 'across' or 'within'. If the latter, only metrics that are set up to work 'within' clusters (rather than 'across' clusterings) are used.